On Mon, Jun 04, 2012 at 12:16:26AM +0200, frantisek holop wrote:
> hmm, on Sun, Jun 03, 2012 at 01:39:18PM +0200, Tobias Ulmer said that
> > > these must be some really nice disks :]
> > >
> > > for example only a 200G slice (also 64k/8k) of music/film/picture
> > > collection (not even full yet) on a notebook disk (5400 RPM) takes ages:
> > >
> > > Filesystem Size Used Avail Capacity iused ifree %iused
> > > Mounted on
> > > /dev/sd0d 217G 153G 63.5G 71% 44815 7197423 1% /data
> > >
> > > $ time sudo fsck -f /dev/sd0d
> > > ** /dev/rsd0d
> > > ** File system is already clean
> > > ** Last Mounted on /data
> > > ** Phase 1 - Check Blocks and Sizes
> > > ** Phase 2 - Check Pathnames
> > > ** Phase 3 - Check Connectivity
> > > ** Phase 4 - Check Reference Counts
> > > ** Phase 5 - Check Cyl groups
> > > 44815 files, 20076091 used, 8329340 free (13748 frags, 1039449 blocks,
> > > 0.0% fragmentation)
> > > 4m58.26s real 0m22.50s user 0m7.28s system
>
> at 71% disk usage having 1% inode usage, would it be a logical
> idea to radically slash the number of inodes, perhaps by 50%, even more?
>
> if i had 50% of the current total inodes, would the fsck time
> be halved? for some reason it seemed logical that
> checking free inodes will be much faster then used ones...
>
> > This comes down to the FFS1 vs FFS2 difference. Newfs will select FFS2
> > for bigger filesystems, reducing fsck times significantly at the expense
> > of more efficient disk space allocation in FFS1.
>
> by efficient disk space allocation you mean fragmentation?
> are there any numbers comparing FFS1 to FFS2 in this regard?
>
> would there be a perceptible (negative) effect of using FFS2 on slices
> smaller than 1TB?
>
> -f
> --
> experience is nothing but a lot of mistakes.
There are two major differences between ffs1 en ffs2
1. ffs2 inodes are twice as big, since the block number sizes have
doubled. This has the consequence that the meta data of a ffs2
filesystem take more space.
2. ffs2 initializes inode blocks on a 'as needed' approach. So on a
typical filesystem, you have far less inode active blocks compared to
the ffs1 situation. (that also explains the much quicker newfs on a
ffs2 filesystem).
Empty inodes do need to be check to see if they are really empty (do
not refer blocks allocated elsewhere), while nonexistent inodes you do
not have to/cannot check. That largely explains the speed difference.
For smaller file systems, using ffs2 can speed fsck up, but you'll
waste some more space on meta data. Note that inode blocks in the ffs2
case always are reserved, the unused inode blocks still take up space.
-Otto
